What's shipping

MCP server: OAuth connector support, scoped permissions, and 4 new tools

• •Tool count is now 24, up from the original 20 — new tools add get_reply_conversation (read a reply thread's actual message content and tags, not just its intent classification), get_variant_performance (per-step, per-A/B-variant send/open/click/reply/bounce breakdown), start_campaign, and review_held_job (approve or reject a held pre-send job)
• •ForgeSend's MCP server can now be added as a real OAuth-based connector — including directly from Claude's own Connectors screen — alongside the existing workspace API key method, so supported AI clients get a real login and consent screen instead of manual key-copying
• •Per-tool permission scoping is now enforced — each of the 24 tools checks the connecting key or token's actual granted scopes, using the same Contacts, Campaigns, Analytics, Sending, Lists, and Webhooks scopes already used elsewhere in Settings, so a connection can be limited to exactly what it needs

Warmup deliverability tools: blacklist monitoring and inbox placement testing

• •A new placement test shows exactly where warmup emails are landing — inbox, spam, promotions, or blocked — not just whether they sent
• •Each placement test includes a content and spam-risk score with a letter grade and specific recommendations for improving placement
• •Warmup inboxes can now be monitored against major email blacklists directly from the warmup dashboard
• •Blacklist status refreshes automatically on a recurring schedule, with a manual refresh available at any time

Self-hosting MVP: Dockerfiles, docker-compose, and BYOK enrichment

• •Self-hosted MVP support is live: Dockerfiles and a docker-compose stack bring up Postgres, Redis, backend, and frontend with one command, plus a bootstrap script that creates the first admin account and workspace
• •BYOK enrichment for the self-hosted Solo tier — bring your own FullEnrich key, billed and used at your own account's rate. This is a single-operator MVP: one shared enrichment key per instance; per-workspace BYOK for multi-client Agency use isn't part of this release yet

In-app searchable Help Center

• •A searchable Help Center is now available inside the dashboard — find answers without leaving the app
• •Help content stays in sync with ForgeSend's documentation automatically, so in-app answers never drift from the public docs
• •Every doc page also renders natively in-app, with the same interactive examples as the public docs site

REST API: rate limiting, Sending guard, and outbound webhooks

• •Per-workspace rate limiting now protects the Contacts, Lists, Campaigns, Analytics, and Sending APIs
• •The Sending API now requires a dedicated sending-scoped API key, workspace- and inbox-scoped, on top of existing authentication
• •A new outbound webhook dispatcher delivers 6 event types with HMAC-signed payloads and an automatic retry queue
• •Webhooks are managed entirely from Settings — register an endpoint, view delivery history, enable, disable, or delete it, no code required

Proactive AI assistant nudges

• •The AI assistant now proactively surfaces nudges instead of waiting to be asked — a nudge check runs automatically on every dashboard visit
• •Nudge cadence is configurable per workspace: every 3 days, weekly, biweekly, or off entirely
• •Each nudge is deduplicated and logged, so the same suggestion doesn't repeat before its cadence window has passed

Live multi-provider enrichment waterfall

• •Lead enrichment now runs through a live multi-provider waterfall spanning 15-20+ data sources, replacing the earlier single-path search and import flow
• •Cheaper sources and cache hits are tried first, with the waterfall falling through to additional providers only when needed to fill in a result
• •Every enriched contact runs through triple email verification — format check, MX record check, and mailbox existence check — before it's usable in a campaign
• •Every provider call and cache hit is still logged to the enrichment ledger with exact cost, so the added provider depth never affects cost transparency

MCP server — connect ForgeSend to AI agents

• •New MCP server exposes ForgeSend to any MCP-compatible AI agent, authenticated with a workspace API key
• •Full campaign tools: create, list, get, update, pause, resume, and delete campaigns, plus add, update, and delete individual sequence steps
• •Lead tools: search, add, update, and delete contacts, and enroll an entire list into a campaign in one call
• •Read-only analytics tools: per-workspace and per-campaign send/open/click/reply/bounce stats, plus daily send volume

Automated email warmup

• •Automated warmup ramps new sending domains safely before campaigns launch, with reply simulation and inbox placement tracking
• •Ramping and Maintenance volume presets, or a custom daily volume — always synced with the inbox's own daily sending cap
• •Inbox health score (DNS, domain age, spam rate, warmup duration) and an inbox-vs-spam-vs-other placement trend graph in the warmup dashboard
• •Custom warmup email templates per inbox, with bulk pause/resume across every warming inbox in a workspace

Workspace control centre with AI-generated summaries

• •Full-screen workspace control centre combining campaign list, inbox health snapshot, and a 7-day send volume chart in one view
• •Alert strip surfaces issues across the workspace that need attention
• •New AI workspace summary turns current workspace state into a plain-language brief, billed against the enrichment ledger
• •Campaign list and inbox health share live data with the rest of the dashboard, so nothing goes stale between views

Per-variant analytics and automatic winner detection

• •Analytics now break results down per A–E variant, alongside existing per-step and per-inbox breakdowns
• •Set a winning variant directly from analytics — a winner badge appears and losing variants are deactivated automatically
• •Variant performance is visible in both the campaign detail workbench and the Jobs tab
• •Every send job carries its variant label through to job status and progress views

Variant dispatch: B–E variants in round-robin rotation

• •Up to five variants (A–E) per step now participate in live send-time selection, not just preview
• •The scheduler rotates active variants round-robin across outbound sends
• •Every send job persists which variant it sent, so results trace back to the exact copy that went out
• •Deactivating a variant removes it from rotation immediately without affecting jobs already sent

Multi-workspace UI

• •Sidebar workspace switcher moves operators between workspaces without leaving the current page
• •Settings workspace tab lists every workspace with rename, switch, and create-new actions
• •Dashboard workspace grid gives an overview of all workspaces before entering the control centre for one
• •Switching or renaming a workspace refreshes the active view immediately

Same-thread sending improvements

• •Follow-up steps that continue the same thread now correctly reuse the prior Message-ID, Gmail thread ID, and subject line
• •Same-thread steps can intentionally leave the subject blank instead of forcing a re-generated line
• •Step 1 remains forced to start a new thread — only later steps can opt into same-thread continuation
• •Reply-style subject handling fixed so continued threads read correctly in the recipient's inbox

Open and click tracking, with a workspace tracking toggle

• •Outbound emails can carry an injected open-tracking pixel and rewritten, tracked links
• •Campaign analytics display real opened and clicked counts and rates alongside existing metrics
• •Tracking is workspace-configurable — turn it on or off from Settings rather than having it forced on
• •Analytics show opened/clicked data only for campaigns sent while tracking was enabled

Campaign duplication and a pre-send review queue

• •Duplicate any campaign to clone its schedule, settings, steps, variants, tags, and sender pool in one action
• •Duplicated campaigns land as drafts with no leads copied over, so testing a variation never touches live contacts
• •New pre-send review queue lets a campaign hold jobs for approval before they go out
• •Jobs workbench exposes held jobs with approve and reject actions next to the rest of the send queue

Sender pools and send policy controls

• •Campaigns can select a specific pool of inboxes to send from, instead of always drawing from every connected inbox
• •Pool-aware readiness checks and preflight scoping confirm the selected pool has capacity before launch
• •New send policy controls skip risky or unverified contacts and set minimum/maximum delay between sends
• •Policy checks run at lead enrollment, enqueue, and dispatch, with status surfaced in the launch and preflight UX

Visual sequence canvas with conditional branching

The campaign builder now includes a full-screen visual canvas powered by ReactFlow. Drag email steps, conditions, wait nodes, and end nodes onto an infinite canvas. Connect them to build branching sequences — route leads down different paths based on whether they replied, opened, or received no response. Canvas state persists per campaign. Apply canvas email steps directly to the live sequence with one click.

ForgeSend Sequence Copilot

Generate complete multi-step cold email sequences using AI. Input your company context, ICP, offer, campaign goal, and tone. The Sequence Copilot produces subject lines, body copy, wait days, and spintax variations pre-applied — spam-scored before you import. Each step is editable before applying to your campaign. Powered by Claude Sonnet.

Reply Box — thread view, intent tagging, and compose reply

The reply inbox is now fully functional. View full conversation threads, compose and send replies without leaving ForgeSend, and tag each reply with intent — Interested, Not Interested, Out of Office, or Unsubscribe. Tagging a lead as Interested or Unsubscribe automatically pauses their sequence. Filter replies by tag, campaign, inbox, and read status.

Campaign sequence builder overhaul — step editing, A/B variants, schedule, and email scoring

Five improvements shipped together: merge tag fallback syntax ({firstName|there}), global suppression list with auto-suppress on unsubscribe detection, full step editing with A/B/C/D/E variant tabs per step, sending schedule with day/hour/timezone/daily cap controls, and a live email scorer — 100-point score across spam word detection, word count, link count, variable coverage, and spintax variation. Preview mode shows exactly how each step renders per lead before sending.

Pre-send guardrail checklist wired into campaign builder

The pre-send guardrail panel is now embedded directly inside the campaign builder. Selecting a campaign immediately fetches its preflight verdict — BLOCKED, WARNING, or READY. The Start button is disabled until all blockers are resolved. A Re-check button lets operators re-fetch without a page refresh. Backend enforcement remains active as the hard guarantee.

Deliverability dashboard with real DNS probes

A dedicated deliverability dashboard now runs live DNS checks — MX, SPF, DKIM, and DMARC — per connected domain. Per-inbox readiness is scored and surfaced with READY/WARNING/BLOCKED verdicts. Workspace-level capacity and domain health checks run on demand.

3-tier email verification with UI badges

Contact lists can now be verified before sending. Verification runs three checks in sequence: format validation, MX record lookup, and live SMTP RCPT probe via raw socket. Each contact receives a verification status badge — Valid, Risky, or Invalid — visible on the list detail page.

Auto-pause on reply and bounce spike

Campaign sequences now pause automatically when a prospect replies. Leads are marked REPLIED and pending send jobs are skipped immediately. Bounce spike detection monitors failure rates per sending job and flags inboxes that exceed safe thresholds.

Real email dispatch via Google and Microsoft OAuth

ForgeSend now sends real emails via Gmail API and Microsoft Graph. Connect Google Workspace and Microsoft 365 accounts directly via OAuth — no SMTP passwords required. SMTP and IMAP inboxes are also supported for any provider via the manual setup flow.

Safety checks and verification infrastructure

Pre-send safety checks added to the sending scheduler. Unsubscribe language detection, daily send limit enforcement, and per-inbox capacity checks run before every batch. The verification service and verify-list endpoint are live.

Analytics and unified reply inbox

Workspace-wide campaign analytics now surface sent, opened, clicked, bounced, and replied counts per campaign. The unified reply inbox aggregates all inbound replies across every connected inbox into a single view. Click any row to mark as read.

Smart sending with inbox rotation

The sending scheduler now distributes outbound sends across all connected inboxes via round-robin rotation. Daily send limits are enforced per inbox. The scheduler runs continuously and processes pending send jobs in order of campaign step and contact priority.

Campaign sequences with spintax

Multi-step campaign sequences are now live. Each step supports spintax — {{variation one|variation two}} syntax — for subject lines and body copy. Custom delays between steps are configurable per step. Spintax renders uniquely per send to improve deliverability. (Note: spintax syntax changed to double braces on 29 July 2026 to disambiguate it from merge-tag fallback syntax — this entry has been updated to reflect current syntax.)

Enrichment waterfall and cost transparency ledger

The first ForgeSend release. Enrichment runs a provider waterfall — cheaper providers first, with cache hit detection before any paid call. Every enrichment attempt — cache hit, provider success, provider failure — is logged to the transparency ledger with exact cost. The ledger is visible per workspace with total balance spent, tasks processed, and cache hit rate.